CENTRE FOR MEDICAL STATISTICS
SEMINAR SERIES on BIOSTATISTICS 2000/2001
NO 9: Wednesday, June 6th, 2001, at 2:30 pm
"Generalising the generalised linear mixed model"
Dr Sophia Rabe-Hesketh
(Institute of Psychiatry,
London)
An extension to generalised linear mixed models allows a large number of
multilevel latent variable models to be fitted to
unbalanced data with mixed responses (requiring different links and
distributions). The linear predictor is simply extended to allow
each random effect to multiply not just a single variable as in random
coefficient models, but to multiply a linear combination of
variables. This is implemented in a Stata program called gllamm which
estimates the models by maximum likelihood. Instead of
assuming multivariate normality of the random effects (or latent
variables), discrete distributions can also be estimated. Examples
of problems with mixed responses that can be tackled with this setup are
multilevel factor and structural equation models for
unbalanced data, multilevel models with errors in covariates and
multilevel models for ordinal and nominal responses, rankings and
multivariate survival data. The class of models will be described and
several examples given. The work was done jointly with Andrew Pickles
and Anders Skrondal.
All welcome!!
Venue:
Room 2.22, Third Floor
The MacKay Building
Keele University
http://www.keele.ac.uk
http://www.keele.ac.uk/depts/ma/seminars/medstats.html
http://www.keele.ac.uk/university/campus/maps/
___________________________________________
Janet Drewery
Secretary
Centre for Medical Statistics
Keele University
Staffordshire ST5 5BG
England
Email: [log in to unmask]
Tel: (01782) 583269
Fax: (01782) 583269/584268
|